29
votes
Using ASCII art to prove that I, not an LLM, wrote something
Link information
This data is scraped automatically and may be incorrect.
- Title
- To Art is Human - proving you wrote that text by making it pretty
- Authors
- Joe Kaufeld,
- Published
- Sep 24 2026
- Word count
- 4151 words
This is a blog post that I spent the last two months working on. I've run into a situation at work where there are several people who aggressively vibe-code all the things, and while I think that LLM usage is helpful in some cases, I don't believe that outsourcing your brain to it has any real usefulness. This post is about the techniques that I've adopted at work, and the history behind them, for proving beyond a shadow of a doubt that the documentation I write is by me and me alone.
I'd love to know your thoughts!
Your mobile-friendly link at the top links to example.com, not the mobile friendly version.
I knew someone was going to point that out before I fixed it
The build that fixed it failed before I noticed, and it's re-running now. >_<
edit: it's fixed now!
If there's one thing I've learned about website design, blog publishing, documentation publishing, and commenting on things it's that the nanosecond I push to production I'll catch a glaring error that I should've caught in one of the 4-5 passes before I finalized and published. Every time. Without fail.
Every single freakin' time.
For me, it’s the same thing but when I’m writing emails. I can’t tell you how many times I just completely fail to spot an error when I’m looking for one, but my eyes lock onto it as soon as I hit Send. As if my subconscious logged it on my Nth re-read and review, but decided to only bring it up once it was too late.
I absolutely love the option to “delay send” emails, I have it switched on for basically every service and platform whenever possible.
I find that the most common thing that happens is that I have accidentally typed a word that has a similar shape to the word that I actually want, and my brain keeps glossing over it. Like I’ll type “expatriation” rather than “expectation” and they’re such a similar word shape that I’ll just… voom.
It’s not just long words either, but I also find that I’m most error prone on phones due to a combination of tiny fonts on my increasingly large screens, the fact that I can’t really blind type like I can on a computer with a keyboard (looking at what I’m actually typing) because I have to stare at the screen to make sure that my thumb is positioned correctly because I slide to type as two thumb typing is really awkward and cramped for me, autocorrect being too aggressive in picking my words, and finally really terrible tools for actually editing the text and correcting it.
I make a ton of errors on here because I’m usually on my phone.
Our brnias are vrey apdet at rarrnageing ltteers eevn if eervy wrod is scrblamed.
Finding errors in texts can be a little difficult for our brains because we're sometimes too good at interpreting words.
Another problem that I have is that I don’t actually type every individual letter in a sentence when I’m typing on a keyboard; what I’m actually doing is performing what is akin to a macro with my fingers, where I know the pattern of the keys that I need to press so I just sort of sent a pulse to my fingers and they tap in a certain order and I get what I need; it honestly feels like what I imagine magic must feel like in systems where you use gestures to cast spells or shape magic. The trouble comes in when I’ve pulsed the wrong word, or when I’m stiff and cramped and my fingers aren’t flexing like they used to. My left index finger is particularly prone to not fully engaging the key or twitching and double engaging a key, or even sometimes engaging the key, but much later than it should have.
Maybe I'm misunderstanding, I'm reading this early morning after a terrible night's sleep.
Are you claiming that llms can't generate ASCII art? I believe they have been able to do this for a while.
I'm claiming that mostly they can't generate something where each line is the same length, and the ASCII art (that is topical and relevant to what you're writing about) is the cherry on top.
(And there's a little bit of yelling at the world thrown in there too.)
If you haven't seen it, Tom7 made a tool to do exactly this: https://youtu.be/Y65FRxE7uMc (about halfway through)
Not intending to rain on your parade or anything, just thought it was relevant.
Never heard of this channel, but also, what the hell, Tom7? What the hell. Very impressive way of tackling the problem.
It does sadden me that we, as creatives, as writers, are spending so much of our collective brain cycles making sure that something we wrote may never be construed as LLM output, when the plain and simple fact of the matter is that you can usually tell if it is by the content not being particularly interesting.